Funny, I've seen people with resistances like what you say (measured them), but for myself, I get like 200K (that's 200,000 Ohms) with wet hands and often over 1M with dry hands....
Of course, it drops below 20K when I have wet, SALTY hands (stupid potato chips). Then again, that's about the worst combination you can achieve.
Funny, I once found myself part of a circuit that I was working on... I was using an old-fashioned JFET as a switch, or I thought I was... Interestingly, it would switch "on" whenever I touched the wire that lead to the "Gate", and switch "off" when I released it.
But then I disconnected my multimeter and the circuit stopped working (who uses a 5Kohm multimeter anyway?). Putting a 5K resistor where the meter had been made it work again...
Sometimes resistance paths make less and less sense (I imagine that it was me, acting as an antenna, that caused the switching).
This is due to what's called "RF grounding". You cannot get a shock from one end of a DC source, no matter what voltage (you can get a small static shock as your potential matches to it, but that's it).
An AC source, however, you're acting as an antenna. Any amount of current you radiate into the air causes current to flow through you, hence the shock. At 60Hz, you're only getting a little wussy shock, it would have been a LOT worse if you'd grounded back into the other end of the socket.
At higher frequencies (think Tesla coils) you actually can radiate most of the current, although at the higher end, the "skin" effect will protect you to some degree (hence Tesla used to stand on insulating platforms, touching a live coil with one hand and fire lightning from his fingers;)
The risetimes of static can generally be handled just fine with good diodes (the high-frequency response can be good, even in a Zener diode, if they are specifically designed for it).
Drivers and receivers hardened against static are actually generally trivial - if you're willing to put Bipolar logic on a CMOS motherboard. Most companies opt for an all-MOS technology these days, but the MOS transistor is really the culprit for ESD-problems (the thin metal-oxide can be vaporised by a very small overvoltage, often in the vicinity of only 12V or so).
The old-fashioned bipolar transistor is faster (better high-frequency response), has a much higher current gain (so it's well suited to drivers anyway), and doesn't rely on a voltage building across an internal capacitor. I used to play with TTL DIP's for fun when I was like 15 - in the worst ESD conditions you can imagine (usually on my bed, with no grounding around, running off a battery-driven regulator). To the best of my knowledge I never popped one. (For the technically-inclined, static is usually low-current, high-voltage. CMOS is damaged by high-voltage, bipolar by high-current. It's an oversimplification, but it's generally a pretty good rule-of-thumb).
I certainly agree with the careful mechanical design thing though, the ungrounded metal chassis on the newer Palm's are really begging for trouble.
I thought Dell was also known for using underrated 145W PSU's in their systems? (Not sure, might have been one of the other prebuilt companies, but I thought it was Dell).
Usually the frills get cut to lower the prices in these systems. Making them crash less is usually more important than ESD-hardening et all.
ESD-protection diodes are cheap (about $0.05 per diode for the biggies) and you only basically need one per IO line reaching the outside world (the average computer has about 100). Plus these can be built on the IC's, although they would lack the massive clamping currents of the discretes (most of these IC's already have a good bit, the serial ports especially actually).
$5 for peace-of-mind? Sounds like a good trade to me.
More often than not, MP3 is implemented in a seperate chip, made by another company (Micronas Intermetall made the first one, I believe, which was used in the original Rio).
Now, it's possible to do in other hardware, like a DSP or (unlikely) a general-purpose CPU. The second is not really going to happen, MP3 and Vorbis decoding take too many cycles for good power-consumption. A DSP chip would be a possibility, since they tend to have much heavier number-crunching abilities. But the dedicated chips are still the most common.
Short of using reprogrammable hardware (like a CPLD/FPGA), I doubt that we'll see upgrades like this for most portables.
But if the Ogg people want to release a full and comprehensive specification (it's an open format, but it's not properly specified, you have to read the sources) then I'm sure somebody would make a core for it (I'm tempted myself, but I never quite managed to figure out enough of the spec to get my software decoder working).
The case you're describing, where a small current flowing through a big resistor generates a large voltage, only happens if you connect the resistor across a constant current source (which, by the way, is essentially a ficticious device), or a current-storage element (inductive load). A voltage-storage element (capacitative load) will just discharge very slowly.
And unless you're suggesting that you're somehow causing a flow of current to exist by some mysterious (magnetic? thermal?) property, which in any case would act more like a voltage source and thus be current-limited by the resistor, a potential-difference on the connectors will have to be drawn THROUGH the resistor, hence the current will be the dependant variable, not the voltage, since the potentials are fixed.
A more likely problem is just a high level of potential difference between the two devices before they are connected (static electricity), so that a high potential is developed, which can quite rapidly pop the insulating layer in a MOSFET if not properly clamped with big diodes.
Sorry, but a surge-suppressor wouldn't do anything in this case. Surge-suppression just makes sure that no really large voltages develop across the power-lines.
What a situation like this calls for is a whole mess of diodes. Probably hi9h-rated zeners. You'd be surprised how few engineers actually design products (especially computer products) with the absolute maximum rated stresses in mind. Just think about how easily you could kill most of the Palm's by dropping them.
Adding big diodes could bleed away such static before it kills the system, although an even better solution would be a hinged door over the socket on the cradle (like on game consoles) with semiconductive plastic in the cradle itself to ground out any loose charge on the palm.
Sorry, you've got that one wrong. RS-232 signals are supposed to have a noise-rejection between like -6 and +6 V. What you're actually getting are charge-pumped devices which increase and invert the output voltages (Maxim's MAX232 series does the job for 0-5V power-supplies, and similar exist for lower supplies, check out their line if you want more information).
But yes, RS-232 are supposed to have rediculous amounts of clamping (the signal lines come in through resistors, and are fed through big clamping diodes). On BOTH sides of the interface chip, they're supposed to fry before they let the connected equipment fry, and they're supposed to be really paranoid about signals anyway.
There ARE no specs. Apparently Vorbis was designed on a newsgroup, and in the sources. They plan to put specs up, the shells of the specs are up, but no specs in them.
Something about they haven't gotten around to it. Mind you, now that 1.0 is out(?) the spec should be finalized, if somebody bothers to write it.
Resolution is 1 packet, which amounts to a few hundred to a few thousand samples. On the other hand, it makes it possible to decode any random packet, as long as you've read the headers, so you could technically decode/edit/reencode and pretend that it was sample-granular. Unlike MP3, Vorbis could actually hide it all for you.
There are 2 or 3 companies that make dedicated hardware mp3 player chips. These companies pay royalties per chip to Fraunhoffer for the license, so the person making use of the chip doesn't have to.
All the manufacturers would love to have a technology they could implement for just the cost of a few engineers to make it go.
What you're referring to is a "Psychoaccoustic Model", and they lie at the heart of any perceptual audio compression. Specifically, there are several theories having to do with human perception of sound (Bark scales and subbands, masking, etc), out of which they build a synthetic "Golden Ear".
Any good encoder will actually optimize as best as it knows how to fit within the limits of the "Golden Ear" and your bandwidth rules. The least important audio data will be the first to go. The amount of "padding" left is determined by the bitrate.
An interesting side-affect is "phase-shifting", where some frequencies will move through time. As long as the amount of motion is smooth, the human ear will not generally perceive it (which is why filters for audio use have such atrocious phase-responses). Mind you, this can mess up if it's not a smooth delay. Digital filters don't usually manage to skew the phase, but some audio compressors do.
I love my Dell XPi. 16MB RAM (I think), P75, 4.3Gb HDD. I even managed to get 640x480x16bit out of it. Lithium battery does 3.5 hours under load, and can stay suspended to memory for about a week.
The best thing is that I got it for about $500 Canadian (what's that, about $300-$350 US if memory serves me) last year. Still does everything I need, especially with the PCMCIA ethernet card, external SCSI CDROM, and ZIP 250 I got for it (I spent more in peripherals for my laptop than on the laptop itself), although the lack of sound is a pain. Having an MP3 player helps that though, I just wish the RIO supported streaming audio.
I posted up above that the lack of a formal spec is what's holding this up. A single dedicated hardware hacker could easily code up something to run on a dedicated DSP, or do the actual VHDL/Verilog/whatever for a dedicated chip pretty easily. Then they could load it into a standard off-the-shelf FPGA chip, so they could just open-source the chip itself.
Just watch how fast the chip is produced in real silicon if the implementation is released under a non-restrictive-enough license.
(I like and use Vorbis enough that, if I weren't so busy, I'd take on this project myself).
Last I heard, the only specs that you could get for Vorbis were the sources... Free or not, you're basically stuck reverse-engineering (very legally, of course, since the head of the project actually suggested it to me;) the source code.
You're ubiquitous hardware devices are the result of about 3 different single chips, which accept MP3 and output digital data. When somebody with the skills gets ahold of a formal spec, then we'll see where Vorbis goes.
Basically, young people have no real rights. They are all disenfranchised and at the whims of old men.
And this is exactly why "hackers" are dangerous. They keep kids down by limiting our access to information so that we can't actually do anything about it (I'm talking the highly-technical type of stuff, like programming, logic design, hardware, etc). School exists to normalize us all, and keep down the radicals, or redirect them into "radical" acts like wearing shirts with a logo on them...
The internet is changing all that. There are now a small few young people who can do a lot of highly-skilled things at young ages, and these people are disgruntled. You get a lot of script-kiddies, but more serious types are showing up, and often enough are a real threat to the establishment, since we form our own views of society before they can make us conform.
So you're seeing a political revolution on the internet (Open-Source comes to mind, you know how the US feels about "communism"), and the corporations want to stop it at all cost (although this has little to do with this law in particular, other than supplying the hysteria on which it is based).
When rapists get so little, but *playing network Doom/Quake/etc* in school will get us 5 to 10.
(Well, ok, I'm in Canada, besides which most of the teachers were playing too.... Some of them were MEAN shots even... I even had one class where the course objective was to "assemble computers from parts, assemble a network, and play network-Doom"...)
Just thinking that parametric model-generating techniques are getting more advanced. How long before the "standard model library" is parametric (so your office-tables can be virtually any shape, for example).
There are already a few programs to do this with human models, although they are admittedly primitive right now (Poser comes to mind). But remember, whenever a precedent for a technology exists, the technology itself is coming right behind it.
I had coordination and motor control problems as a child. I played Star Control 1 till I could beat the "Awesome" computer with only one ship (either the Arilou or the Myrnmyrrm for those who are counting, or the Spathi on the other side). Now I don't remember the last time I've been beaten by a human opponent in one of the SC games when they didn't have three or more times the fleet-size as me;)
Seriously though, if you practice, your coordination gets better. Sit down and wire up a 20000+-wire breadboarded *something* (it was a neural net in my case) and you'll get a *lot* better at wiring. Nowadays I've got excellent dexterity and okay hand-eye coordination. I really only play about 3 hours of games a week (well, lump it together, its more like a solid week once a month or two), and code (and hardware equivalent) the rest, but still, I say it helped me.
I have a box of about 50+ 5.25 drives, although I admit I'm missing the 8" drives (I should visit ebay...). Got a card-reader though (standard RS-232C), and even a 3.25" drive (yes, that's not a typo), although that one is part of a word-processor system.
Then again, I'm a bit of an antique collector. I've got literally dozens of 1980's computers. But there will always be collectors like me who can COPY these formats (does anybody need a few thousand punch-cards converted to ASCII-text and burned on a CD?). So I don't think there's much of a problem here. And the more widely used a format (think Zip vs Orb vs plain CDR) the more of us will have one around (or a box of 50+ of them).
I'm just worried that one of these "Globalization" agreements will extend foreign laws into other countries. Up here in Canada, we lack a DMCA, and I also seem to recall that "keeping a backup copy for archival use" is considered "fair use". The way the globalization stuff (FTAA, and its ilk) is going, I'm worried that we won't need a DMCA, US companies will just be able to sue us without one.
Slashdot Mirror
Of course, it drops below 20K when I have wet, SALTY hands (stupid potato chips). Then again, that's about the worst combination you can achieve.
But then I disconnected my multimeter and the circuit stopped working (who uses a 5Kohm multimeter anyway?). Putting a 5K resistor where the meter had been made it work again...
Sometimes resistance paths make less and less sense (I imagine that it was me, acting as an antenna, that caused the switching).
An AC source, however, you're acting as an antenna. Any amount of current you radiate into the air causes current to flow through you, hence the shock. At 60Hz, you're only getting a little wussy shock, it would have been a LOT worse if you'd grounded back into the other end of the socket.
At higher frequencies (think Tesla coils) you actually can radiate most of the current, although at the higher end, the "skin" effect will protect you to some degree (hence Tesla used to stand on insulating platforms, touching a live coil with one hand and fire lightning from his fingers ;)
Drivers and receivers hardened against static are actually generally trivial - if you're willing to put Bipolar logic on a CMOS motherboard. Most companies opt for an all-MOS technology these days, but the MOS transistor is really the culprit for ESD-problems (the thin metal-oxide can be vaporised by a very small overvoltage, often in the vicinity of only 12V or so).
The old-fashioned bipolar transistor is faster (better high-frequency response), has a much higher current gain (so it's well suited to drivers anyway), and doesn't rely on a voltage building across an internal capacitor. I used to play with TTL DIP's for fun when I was like 15 - in the worst ESD conditions you can imagine (usually on my bed, with no grounding around, running off a battery-driven regulator). To the best of my knowledge I never popped one. (For the technically-inclined, static is usually low-current, high-voltage. CMOS is damaged by high-voltage, bipolar by high-current. It's an oversimplification, but it's generally a pretty good rule-of-thumb).
I certainly agree with the careful mechanical design thing though, the ungrounded metal chassis on the newer Palm's are really begging for trouble.
Usually the frills get cut to lower the prices in these systems. Making them crash less is usually more important than ESD-hardening et all.
$5 for peace-of-mind? Sounds like a good trade to me.
Now, it's possible to do in other hardware, like a DSP or (unlikely) a general-purpose CPU. The second is not really going to happen, MP3 and Vorbis decoding take too many cycles for good power-consumption. A DSP chip would be a possibility, since they tend to have much heavier number-crunching abilities. But the dedicated chips are still the most common.
Short of using reprogrammable hardware (like a CPLD/FPGA), I doubt that we'll see upgrades like this for most portables.
But if the Ogg people want to release a full and comprehensive specification (it's an open format, but it's not properly specified, you have to read the sources) then I'm sure somebody would make a core for it (I'm tempted myself, but I never quite managed to figure out enough of the spec to get my software decoder working).
You're right, of course, which is what I was saying. Perhaps you meant to reply to the parent to my message?
And unless you're suggesting that you're somehow causing a flow of current to exist by some mysterious (magnetic? thermal?) property, which in any case would act more like a voltage source and thus be current-limited by the resistor, a potential-difference on the connectors will have to be drawn THROUGH the resistor, hence the current will be the dependant variable, not the voltage, since the potentials are fixed.
A more likely problem is just a high level of potential difference between the two devices before they are connected (static electricity), so that a high potential is developed, which can quite rapidly pop the insulating layer in a MOSFET if not properly clamped with big diodes.
What a situation like this calls for is a whole mess of diodes. Probably hi9h-rated zeners. You'd be surprised how few engineers actually design products (especially computer products) with the absolute maximum rated stresses in mind. Just think about how easily you could kill most of the Palm's by dropping them.
Adding big diodes could bleed away such static before it kills the system, although an even better solution would be a hinged door over the socket on the cradle (like on game consoles) with semiconductive plastic in the cradle itself to ground out any loose charge on the palm.
But yes, RS-232 are supposed to have rediculous amounts of clamping (the signal lines come in through resistors, and are fed through big clamping diodes). On BOTH sides of the interface chip, they're supposed to fry before they let the connected equipment fry, and they're supposed to be really paranoid about signals anyway.
Something about they haven't gotten around to it. Mind you, now that 1.0 is out(?) the spec should be finalized, if somebody bothers to write it.
Resolution is 1 packet, which amounts to a few hundred to a few thousand samples. On the other hand, it makes it possible to decode any random packet, as long as you've read the headers, so you could technically decode/edit/reencode and pretend that it was sample-granular. Unlike MP3, Vorbis could actually hide it all for you.
All the manufacturers would love to have a technology they could implement for just the cost of a few engineers to make it go.
Any good encoder will actually optimize as best as it knows how to fit within the limits of the "Golden Ear" and your bandwidth rules. The least important audio data will be the first to go. The amount of "padding" left is determined by the bitrate.
An interesting side-affect is "phase-shifting", where some frequencies will move through time. As long as the amount of motion is smooth, the human ear will not generally perceive it (which is why filters for audio use have such atrocious phase-responses). Mind you, this can mess up if it's not a smooth delay. Digital filters don't usually manage to skew the phase, but some audio compressors do.
The best thing is that I got it for about $500 Canadian (what's that, about $300-$350 US if memory serves me) last year. Still does everything I need, especially with the PCMCIA ethernet card, external SCSI CDROM, and ZIP 250 I got for it (I spent more in peripherals for my laptop than on the laptop itself), although the lack of sound is a pain. Having an MP3 player helps that though, I just wish the RIO supported streaming audio.
Just watch how fast the chip is produced in real silicon if the implementation is released under a non-restrictive-enough license.
(I like and use Vorbis enough that, if I weren't so busy, I'd take on this project myself).
You're ubiquitous hardware devices are the result of about 3 different single chips, which accept MP3 and output digital data. When somebody with the skills gets ahold of a formal spec, then we'll see where Vorbis goes.
And this is exactly why "hackers" are dangerous. They keep kids down by limiting our access to information so that we can't actually do anything about it (I'm talking the highly-technical type of stuff, like programming, logic design, hardware, etc). School exists to normalize us all, and keep down the radicals, or redirect them into "radical" acts like wearing shirts with a logo on them...
The internet is changing all that. There are now a small few young people who can do a lot of highly-skilled things at young ages, and these people are disgruntled. You get a lot of script-kiddies, but more serious types are showing up, and often enough are a real threat to the establishment, since we form our own views of society before they can make us conform.
So you're seeing a political revolution on the internet (Open-Source comes to mind, you know how the US feels about "communism"), and the corporations want to stop it at all cost (although this has little to do with this law in particular, other than supplying the hysteria on which it is based).
Then again, maybe I'm just jaded and paranoid.
I mean to say, I played more than my share of Doom on school machines. I'm a big criminal now, huh?
(Well, ok, I'm in Canada, besides which most of the teachers were playing too.... Some of them were MEAN shots even... I even had one class where the course objective was to "assemble computers from parts, assemble a network, and play network-Doom"...)
There are already a few programs to do this with human models, although they are admittedly primitive right now (Poser comes to mind). But remember, whenever a precedent for a technology exists, the technology itself is coming right behind it.
Seriously though, if you practice, your coordination gets better. Sit down and wire up a 20000+-wire breadboarded *something* (it was a neural net in my case) and you'll get a *lot* better at wiring. Nowadays I've got excellent dexterity and okay hand-eye coordination. I really only play about 3 hours of games a week (well, lump it together, its more like a solid week once a month or two), and code (and hardware equivalent) the rest, but still, I say it helped me.
Then again, I'm a bit of an antique collector. I've got literally dozens of 1980's computers. But there will always be collectors like me who can COPY these formats (does anybody need a few thousand punch-cards converted to ASCII-text and burned on a CD?). So I don't think there's much of a problem here. And the more widely used a format (think Zip vs Orb vs plain CDR) the more of us will have one around (or a box of 50+ of them).
Scary thought.